|
|
|
| Study of Estimation Performance of Optimally Weighted Stochastic Pooling Networks |
| JING Wentenga, HAN Boa, GENG Jinhuaa, XU Liyanb, DUAN Fabinga
|
| a.Institute of Complexity Science; b.School of Electronic Information, Qingdao University, Qingdao 266071, China |
|
|
|
Abstract In this paper, the optimally weighted stochastic pooling network is investigated for the theoretical and experimental analyses of the random parameter estimation. The stochastic pooling network is first optimized by the random noise components, and then improved by the linear optimum weight coefficients. The theoretical expressions of the optimum weight vector and the mean square error of the stochastic pooling network with an arbitrary number of nodes are deduced. In practice, since the statistical information of parameter and background noise is often unknown, the approximation estimation algorithms of the optimum weight vector and the mean square error are presented and based on the observations. Theoretical and experimental results both verify the optimization ability of random noise, and show the outstanding estimation performance of the optimally weighted stochastic pooling network.
|
|
Received: 30 July 2018
Published: 31 January 2019
|
|
|
|
|
|
|
|
|
